Tools
Document layout
Upload a PDF and see how it is read: every word rebuilt in its place, as structured HTML with headings, paragraphs and tables – and the original beside it.
- HTML
- Tables
Nothing is kept
The PDF goes to the mahevi server, is read there and forgotten right away – no copy, no history. PDFs with a text layer (from Word, a browser …) are recognised completely. Scanned or photographed pages stay empty, because no text recognition runs here.
Rules, AI and layout models
This page runs the rule-based recognition: it derives columns, headings, lists and tables from the position and font of the words and reads clearly set documents reliably. For certain document types – invoices, forms, bank statements – it needs calibrating to their structure, and that is not activated on this page. Complex layouts such as invoices with line-item tables, multi-column contracts or CVs are ordered far better by an AI pass and trained layout models (Azure Document Intelligence): they redetermine the reading order, kind and structure of the blocks. Both cost per page or token, so here they are only shown as buttons.
Daily quota
How many documents you can have recognised a day is shown right above the upload. Without signing in it is fewer than with an account. Once the day is used up, it works again the next day.
Recognise a PDF
Pick a PDF and click “Recognise”. The buttons above switch between rebuild, structure and original.